PACHBOARD
X. Pouwels
7-12-2021
Aim (grant application)
The aim of this project is to develop a generic dashboard, e.g. a R shiny application to 1) inspect model inputs and outputs, 2) visualise the original inputs and outputs, 3) investigate the relationship between model inputs and outputs through metamodelling and data visualisation methods, 4) save the performed analyses.
For info
Package with supporting functions can be found here: https://github.com/Xa4P/pacheck.
This App should focus on validation of the health economic model (using the probabilistic inputs/ outputs).
Structure shiny app
The envisoned R Shiny app will have the following tabs. I’ve developed some functions already to implement. Currently, I’ve only considered 2 strategies (“intervention” (_in) and “comparator” (_comp)). What is shown in this document is based on the probabilistic analysis of a (toy) 3-states Health state transition model (Progression-free (PF), Progressed disease(PD), Dead(D)) depicted in Figure 1. The intervention is only having an effect on the probability of progression, and incurs costs in the Progression-Free health state.
Voor Karel: elke cijfer is een aparte tab van de app.
1. Upload of original health economic model inputs and outputs
Content:
- Welcome message
- Instructions
- Upload 1 file with inputs and outputs (or 2? inputs and outputs separately?)
- Select variables representing total costs and effects for each strategy (to calculate increments and (incremental) net benefits)
df_pa <- calculate_nb(df = df_pa,
e_int = "QALY_int",
e_comp = "QALY_comp",
c_int = "Costs_int",
c_comp = "Costs_comp",
wtp = 80000)# calculate net benefits2. Investigate model inputs and outputs
a. Summary statistics of (user-selected) model inputs and outputs
To examine whether cost inputs are always positive for instance
df <- generate_sum_stats(df_pa)
kable(df)| Parameter | Mean | SD | Percentile_2.5th | Percentile_97.5th | Minimum | Maximum |
|---|---|---|---|---|---|---|
| p_pfspd | 0.150 | 0.035 | 0.088 | 0.226 | 0.049 | 0.316 |
| p_pfsd | 0.100 | 0.030 | 0.049 | 0.165 | 0.023 | 0.244 |
| p_pdd | 0.201 | 0.040 | 0.129 | 0.283 | 0.073 | 0.366 |
| p_dd | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 1.000 |
| rr | 0.753 | 0.068 | 0.630 | 0.895 | 0.535 | 1.033 |
| u_pfs | 0.816 | 0.065 | 0.675 | 0.924 | 0.491 | 0.974 |
| u_pd | 0.587 | 0.103 | 0.382 | 0.776 | 0.207 | 0.898 |
| u_d | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| c_pfs | 998.950 | 202.516 | 603.870 | 1392.049 | 202.083 | 1771.015 |
| c_pd | 1993.689 | 401.306 | 1203.104 | 2782.842 | 388.754 | 3707.492 |
| c_d | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| c_thx | 10000.807 | 98.670 | 9807.326 | 10194.863 | 9632.935 | 10341.933 |
| QALY_comp | 4.005 | 0.743 | 2.727 | 5.644 | 1.939 | 7.889 |
| QALY_int | 4.309 | 0.824 | 2.894 | 6.105 | 2.104 | 8.433 |
| Costs_comp | 7223.596 | 1593.274 | 4473.077 | 10703.240 | 2425.117 | 17054.204 |
| Costs_int | 38841.599 | 7233.474 | 26434.052 | 54575.070 | 19181.834 | 78244.121 |
| Inc_QALY | 0.304 | 0.170 | 0.050 | 0.703 | -0.291 | 1.562 |
| Inc_Costs | 31618.003 | 6576.688 | 20379.534 | 46057.098 | 15154.291 | 66447.151 |
| NMB_int | 305872.059 | 60085.800 | 202019.061 | 436642.750 | 148476.152 | 609534.328 |
| NMB_comp | 313195.478 | 58530.830 | 212540.838 | 442388.664 | 151753.123 | 614042.727 |
| iNMB | -7323.419 | 10571.658 | -24480.486 | 16838.699 | -52000.775 | 61320.703 |
| NHB_int | 3.823 | 0.751 | 2.525 | 5.458 | 1.856 | 7.619 |
| NHB_comp | 3.915 | 0.732 | 2.657 | 5.530 | 1.897 | 7.676 |
| iNHB | -0.092 | 0.132 | -0.306 | 0.210 | -0.650 | 0.767 |
rm(df)b. Correlation matrix inputs (and outputs)
Question Karel:
- kan dit met kleuren om de sterke van de associatie te onderbouwen?
generate_cor(df_pa)## Warning in cor(df): the standard deviation is zero## p_pfspd p_pfsd p_pdd p_dd rr
## p_pfspd 1.000000000 -0.1390333051 -0.010495755 NA 0.013176795
## p_pfsd -0.139033305 1.0000000000 -0.010816151 NA 0.004160546
## p_pdd -0.010495755 -0.0108161511 1.000000000 NA -0.009214208
## p_dd NA NA NA 1 NA
## rr 0.013176795 0.0041605456 -0.009214208 NA 1.000000000
## u_pfs -0.010233113 -0.0139012374 0.006358543 NA -0.010099636
## u_pd 0.019283717 -0.0056315922 -0.002787892 NA -0.010150340
## u_d NA NA NA NA NA
## c_pfs 0.005755797 0.0005144779 -0.015172213 NA -0.001911507
## c_pd -0.001879029 -0.0074655735 -0.020431937 NA -0.014541343
## c_d NA NA NA NA NA
## c_thx -0.001009073 -0.0236999322 0.006581438 NA -0.012871127
## QALY_comp -0.253588264 -0.6735920331 -0.384481458 NA -0.006767802
## QALY_int -0.224667628 -0.7439538243 -0.300979631 NA -0.122611874
## Costs_comp 0.119961253 -0.5263861075 -0.478023551 NA -0.005342058
## Costs_int -0.510900215 -0.7102236038 -0.075277748 NA -0.223503355
## Inc_QALY 0.018834190 -0.6615605116 0.220270065 NA -0.563306427
## Inc_Costs -0.590983549 -0.6536278325 0.033010989 NA -0.244529505
## NMB_int -0.184925056 -0.7305164562 -0.321071795 NA -0.107582111
## NMB_comp -0.260628921 -0.6692910076 -0.377192915 NA -0.006723138
## iNMB 0.391940138 -0.4464301719 0.263493089 NA -0.574237852
## NHB_int -0.184925056 -0.7305164562 -0.321071795 NA -0.107582111
## NHB_comp -0.260628921 -0.6692910076 -0.377192915 NA -0.006723138
## iNHB 0.391940138 -0.4464301719 0.263493089 NA -0.574237852
## u_pfs u_pd u_d c_pfs c_pd c_d
## p_pfspd -0.010233113 0.019283717 NA 0.0057557965 -0.001879029 NA
## p_pfsd -0.013901237 -0.005631592 NA 0.0005144779 -0.007465573 NA
## p_pdd 0.006358543 -0.002787892 NA -0.0151722129 -0.020431937 NA
## p_dd NA NA NA NA NA NA
## rr -0.010099636 -0.010150340 NA -0.0019115067 -0.014541343 NA
## u_pfs 1.000000000 0.007466828 NA -0.0017889376 -0.015444679 NA
## u_pd 0.007466828 1.000000000 NA 0.0046724428 0.012595157 NA
## u_d NA NA 1 NA NA NA
## c_pfs -0.001788938 0.004672443 NA 1.0000000000 -0.000232108 NA
## c_pd -0.015444679 0.012595157 NA -0.0002321080 1.000000000 NA
## c_d NA NA NA NA NA 1
## c_thx -0.008116731 0.007440294 NA 0.0123971920 0.011433502 NA
## QALY_comp 0.269082381 0.382315481 NA 0.0021434831 0.013324345 NA
## QALY_int 0.292116700 0.306768756 NA 0.0004196619 0.012716009 NA
## Costs_comp -0.004667026 0.015616730 NA 0.3427335816 0.585434582 NA
## Costs_int 0.016651801 -0.003383216 NA 0.0858487599 0.121224827 NA
## Inc_QALY 0.239747594 -0.182842530 NA -0.0073115683 0.003416044 NA
## Inc_Costs 0.019445382 -0.007504406 NA 0.0113911703 -0.008496863 NA
## NMB_int 0.318408080 0.336891345 NA -0.0098746547 -0.000645991 NA
## NMB_comp 0.273215262 0.387581921 NA -0.0071541959 -0.002413475 NA
## iNMB 0.297047822 -0.231099574 NA -0.0165144854 0.009690798 NA
## NHB_int 0.318408080 0.336891345 NA -0.0098746547 -0.000645991 NA
## NHB_comp 0.273215262 0.387581921 NA -0.0071541959 -0.002413475 NA
## iNHB 0.297047822 -0.231099574 NA -0.0165144854 0.009690798 NA
## c_thx QALY_comp QALY_int Costs_comp Costs_int
## p_pfspd -0.001009073 -0.253588264 -0.2246676276 0.119961253 -0.510900215
## p_pfsd -0.023699932 -0.673592033 -0.7439538243 -0.526386108 -0.710223604
## p_pdd 0.006581438 -0.384481458 -0.3009796313 -0.478023551 -0.075277748
## p_dd NA NA NA NA NA
## rr -0.012871127 -0.006767802 -0.1226118737 -0.005342058 -0.223503355
## u_pfs -0.008116731 0.269082381 0.2921167001 -0.004667026 0.016651801
## u_pd 0.007440294 0.382315481 0.3067687559 0.015616730 -0.003383216
## u_d NA NA NA NA NA
## c_pfs 0.012397192 0.002143483 0.0004196619 0.342733582 0.085848760
## c_pd 0.011433502 0.013324345 0.0127160085 0.585434582 0.121224827
## c_d NA NA NA NA NA
## c_thx 1.000000000 0.013019157 0.0154765010 0.021052846 0.065247154
## QALY_comp 0.013019157 1.000000000 0.9816698273 0.556292870 0.765003861
## QALY_int 0.015476501 0.981669827 1.0000000000 0.553244013 0.823974129
## Costs_comp 0.021052846 0.556292870 0.5532440127 1.000000000 0.503641319
## Costs_int 0.065247154 0.765003861 0.8239741290 0.503641319 1.000000000
## Inc_QALY 0.018092244 0.388489874 0.5569880350 0.250673380 0.650098971
## Inc_Costs 0.066662827 0.706633521 0.7722315156 0.311676945 0.977853128
## NMB_int 0.009120812 0.984664168 0.9976707214 0.546203013 0.783403016
## NMB_comp 0.012639891 0.999744127 0.9812240653 0.537353311 0.762682825
## iNMB -0.018142098 0.061341401 0.2378041368 0.129337279 0.229948586
## NHB_int 0.009120812 0.984664168 0.9976707214 0.546203013 0.783403016
## NHB_comp 0.012639891 0.999744127 0.9812240653 0.537353311 0.762682825
## iNHB -0.018142098 0.061341401 0.2378041368 0.129337279 0.229948586
## Inc_QALY Inc_Costs NMB_int NMB_comp iNMB
## p_pfspd 0.018834190 -0.590983549 -0.184925056 -0.260628921 0.391940138
## p_pfsd -0.661560512 -0.653627832 -0.730516456 -0.669291008 -0.446430172
## p_pdd 0.220270065 0.033010989 -0.321071795 -0.377192915 0.263493089
## p_dd NA NA NA NA NA
## rr -0.563306427 -0.244529505 -0.107582111 -0.006723138 -0.574237852
## u_pfs 0.239747594 0.019445382 0.318408080 0.273215262 0.297047822
## u_pd -0.182842530 -0.007504406 0.336891345 0.387581921 -0.231099574
## u_d NA NA NA NA NA
## c_pfs -0.007311568 0.011391170 -0.009874655 -0.007154196 -0.016514485
## c_pd 0.003416044 -0.008496863 -0.000645991 -0.002413475 0.009690798
## c_d NA NA NA NA NA
## c_thx 0.018092244 0.066662827 0.009120812 0.012639891 -0.018142098
## QALY_comp 0.388489874 0.706633521 0.984664168 0.999744127 0.061341401
## QALY_int 0.556988035 0.772231516 0.997670721 0.981224065 0.237804137
## Costs_comp 0.250673380 0.311676945 0.546203013 0.537353311 0.129337279
## Costs_int 0.650098971 0.977853128 0.783403016 0.762682825 0.229948586
## Inc_QALY 1.000000000 0.654293217 0.532678288 0.387449729 0.882420376
## Inc_Costs 0.654293217 1.000000000 0.729314498 0.708669007 0.221579204
## NMB_int 0.532678288 0.729314498 1.000000000 0.984454647 0.233156612
## NMB_comp 0.387449729 0.708669007 0.984454647 1.000000000 0.058733889
## iNMB 0.882420376 0.221579204 0.233156612 0.058733889 1.000000000
## NHB_int 0.532678288 0.729314498 1.000000000 0.984454647 0.233156612
## NHB_comp 0.387449729 0.708669007 0.984454647 1.000000000 0.058733889
## iNHB 0.882420376 0.221579204 0.233156612 0.058733889 1.000000000
## NHB_int NHB_comp iNHB
## p_pfspd -0.184925056 -0.260628921 0.391940138
## p_pfsd -0.730516456 -0.669291008 -0.446430172
## p_pdd -0.321071795 -0.377192915 0.263493089
## p_dd NA NA NA
## rr -0.107582111 -0.006723138 -0.574237852
## u_pfs 0.318408080 0.273215262 0.297047822
## u_pd 0.336891345 0.387581921 -0.231099574
## u_d NA NA NA
## c_pfs -0.009874655 -0.007154196 -0.016514485
## c_pd -0.000645991 -0.002413475 0.009690798
## c_d NA NA NA
## c_thx 0.009120812 0.012639891 -0.018142098
## QALY_comp 0.984664168 0.999744127 0.061341401
## QALY_int 0.997670721 0.981224065 0.237804137
## Costs_comp 0.546203013 0.537353311 0.129337279
## Costs_int 0.783403016 0.762682825 0.229948586
## Inc_QALY 0.532678288 0.387449729 0.882420376
## Inc_Costs 0.729314498 0.708669007 0.221579204
## NMB_int 1.000000000 0.984454647 0.233156612
## NMB_comp 0.984454647 1.000000000 0.058733889
## iNMB 0.233156612 0.058733889 1.000000000
## NHB_int 1.000000000 0.984454647 0.233156612
## NHB_comp 0.984454647 1.000000000 0.058733889
## iNHB 0.233156612 0.058733889 1.000000000c. Histogram, density distribution of (user-selected) model inputs and outputs
To visually investigate the parameter distributions
Single parameter
p_1 <- vis_1_param(df = df_pa,
param = "u_pfs",
binwidth = NULL,
type = "histogram",
dist = c("norm", "beta", "gamma", "lnorm"))## Loading required package: fitdistrplus## Loading required package: MASS##
## Attaching package: 'MASS'## The following object is masked from 'package:plotly':
##
## select## Loading required package: survivalp_1## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
paste("Probability to be in user-defined range: ",
check_range(df_pa,
outcome = "u_pfs",
min_val = 0.77,
max_val = 0.80
), "%") # add this to plot? + lines of min/ max on plot?## [1] "Probability to be in user-defined range: 14 %"Two parameters
p_2p <- vis_2_params(df = df_pa,
param_1 = "u_pfs",
param_2 = "u_pd",
slope = 1,
check = "param_2 > param_1")## [1] "P(TRUE): 3 %"p_2p
c. Possibility to fit user-selected distributions (beta, gamma, lognormal, normal) + visualisation + parameters of the fitted distribution.
To cross check with parameters reported in documentation/report/ journal article, as an implementation check
- To do XP: Add probabilistic mean value per distribution
p_2 <- vis_1_param(df = df_pa,
param = "u_pfs",
binwidth = NULL,
type = "density",
dist = c("norm", "beta", "gamma", "lnorm"),
user_dist = "beta",
user_param_1 = 0.8,
user_param_2 = 0.2,
user_mean = 0.75)
p_2## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Distributions’ parameters & statistical fit
l_dist <- fit_dist(df = df_pa,
param = "u_pfs",
dist = c("norm", "beta", "gamma", "lnorm"))
l_dist[[1]]## Distribution AIC
## 1 norm -26434
## 2 beta -27051
## 3 gamma -26080
## 4 lnorm -25867l_dist[[2]]## Distribution Name_param_1 Value_param_1 Name_param_2 Value_param_2
## 1 norm mean 0.82 sd 0.06
## 2 beta shape1 28.78 shape2 6.47
## 3 gamma shape 153.84 rate 188.42
## 4 lnorm meanlog -0.21 sdlog 0.083. Investigate model outputs
a. Incremental cost-effectiveness plane & summary
Questions met Karel te bespreken:
- Interactive plot where you can click on a dot, or select some dots and see which input parameters combination has led to these outputs –> ook mogelijkheid om een bepaalde ‘gebied’ te selecteren (invoeren) en te zien welke inputs daarbij behoren.
- Combineren van de plot hierbeneden met de histogrammen van de waarden op de x en y as, of als niet overzichtelijk: met ticks op de axes als er een datapunt is.
p_3 <- plot_ice(df = df_pa,
param_1 = "Inc_QALY",
param_2 = "Inc_Costs",
wtp = 80000)## Loading required package: scalesp_3
p_3_interact <- ggplotly(p_3)
p_3_interactsummary_ice(df_pa,
"Inc_QALY",
"Inc_Costs")## Quadrant Percentage
## 1 NorthEast (more effective, more expensive) 100%
## 2 SouthEast (more effective, less expensive) 0%
## 3 NorthWest (less effective, more expensive) 0%
## 4 SouthWest (less effective, less expensive) 0%- Plot hierboven koppelen aan een graphs van 1 en 2 input parameters. Al gedaan in tentative shiny app voor onderwijs: selectie van punten in de ‘input graph’ wordt gehighlight in andere plot. Voorbeeld, laatste tabs van volgende shiny app:
shiny::runGitHub("Teaching", "Xa4P", subdir = "Basics/shiny_app_cea/", ref = "main").
b. Cost-effectiveness acceptability curve.
df_ceac <- calculate_ceac(df = df_pa,
e_int = "QALY_int",
e_comp = "QALY_comp",
c_int = "Costs_int",
c_comp = "Costs_comp")
plot_ceac(df = df_ceac,
wtp = "WTP_threshold")## Loading required package: reshape2
df_ceac## WTP_threshold Prob_int Prob_comp
## 1 0 0.0000 1.0000
## 2 1000 0.0000 1.0000
## 3 2000 0.0000 1.0000
## 4 3000 0.0000 1.0000
## 5 4000 0.0000 1.0000
## 6 5000 0.0000 1.0000
## 7 6000 0.0000 1.0000
## 8 7000 0.0000 1.0000
## 9 8000 0.0000 1.0000
## 10 9000 0.0000 1.0000
## 11 10000 0.0000 1.0000
## 12 11000 0.0000 1.0000
## 13 12000 0.0000 1.0000
## 14 13000 0.0000 1.0000
## 15 14000 0.0000 1.0000
## 16 15000 0.0000 1.0000
## 17 16000 0.0000 1.0000
## 18 17000 0.0000 1.0000
## 19 18000 0.0000 1.0000
## 20 19000 0.0000 1.0000
## 21 20000 0.0000 1.0000
## 22 21000 0.0000 1.0000
## 23 22000 0.0000 1.0000
## 24 23000 0.0000 1.0000
## 25 24000 0.0000 1.0000
## 26 25000 0.0000 1.0000
## 27 26000 0.0000 1.0000
## 28 27000 0.0000 1.0000
## 29 28000 0.0000 1.0000
## 30 29000 0.0000 1.0000
## 31 30000 0.0000 1.0000
## 32 31000 0.0000 1.0000
## 33 32000 0.0000 1.0000
## 34 33000 0.0000 1.0000
## 35 34000 0.0000 1.0000
## 36 35000 0.0000 1.0000
## 37 36000 0.0000 1.0000
## 38 37000 0.0000 1.0000
## 39 38000 0.0000 1.0000
## 40 39000 0.0002 0.9998
## 41 40000 0.0007 0.9993
## 42 41000 0.0017 0.9983
## 43 42000 0.0022 0.9978
## 44 43000 0.0030 0.9970
## 45 44000 0.0040 0.9960
## 46 45000 0.0051 0.9949
## 47 46000 0.0061 0.9939
## 48 47000 0.0074 0.9926
## 49 48000 0.0086 0.9914
## 50 49000 0.0099 0.9901
## 51 50000 0.0118 0.9882
## 52 51000 0.0142 0.9858
## 53 52000 0.0170 0.9830
## 54 53000 0.0194 0.9806
## 55 54000 0.0235 0.9765
## 56 55000 0.0274 0.9726
## 57 56000 0.0314 0.9686
## 58 57000 0.0352 0.9648
## 59 58000 0.0390 0.9610
## 60 59000 0.0440 0.9560
## 61 60000 0.0498 0.9502
## 62 61000 0.0552 0.9448
## 63 62000 0.0608 0.9392
## 64 63000 0.0691 0.9309
## 65 64000 0.0756 0.9244
## 66 65000 0.0825 0.9175
## 67 66000 0.0889 0.9111
## 68 67000 0.0963 0.9037
## 69 68000 0.1045 0.8955
## 70 69000 0.1142 0.8858
## 71 70000 0.1243 0.8757
## 72 71000 0.1330 0.8670
## 73 72000 0.1414 0.8586
## 74 73000 0.1507 0.8493
## 75 74000 0.1579 0.8421
## 76 75000 0.1680 0.8320
## 77 76000 0.1763 0.8237
## 78 77000 0.1856 0.8144
## 79 78000 0.1964 0.8036
## 80 79000 0.2046 0.7954
## 81 80000 0.2125 0.7875
## 82 81000 0.2233 0.7767
## 83 82000 0.2334 0.7666
## 84 83000 0.2427 0.7573
## 85 84000 0.2531 0.7469
## 86 85000 0.2621 0.7379
## 87 86000 0.2728 0.7272
## 88 87000 0.2828 0.7172
## 89 88000 0.2917 0.7083
## 90 89000 0.3021 0.6979
## 91 90000 0.3124 0.6876
## 92 91000 0.3216 0.6784
## 93 92000 0.3325 0.6675
## 94 93000 0.3420 0.6580
## 95 94000 0.3510 0.6490
## 96 95000 0.3587 0.6413
## 97 96000 0.3685 0.6315
## 98 97000 0.3792 0.6208
## 99 98000 0.3887 0.6113
## 100 99000 0.3966 0.6034
## 101 100000 0.4054 0.5946c. Histogram and density distribution of total and incremental costs and effects.
Can use the function above!
###d. Convergence graph of outcomes
plot_convergence(df = df_pa,
outcome = "iNMB"
)
4. Investigate relation between inputs and outputs
Single
- To do XP: Add R2
lm_rr <- fit_lm_metamodel(df = df_pa,
x = "rr",
y = "iNMB")
lm_pred <- unlist(predict(lm_rr, data.frame(rr = df_pa$rr)))
df_obs_pred <- data.frame(
Values = df_pa$rr,
Observed = df_pa$iNMB,
Predicted = lm_pred
)
ggplot(data = df_obs_pred, aes(x = Values, y = Observed)) +
geom_point(shape = 1, colour = "lightgrey") +
geom_smooth(method = "lm") +
theme_bw()## `geom_smooth()` using formula 'y ~ x'
plot(lm_rr)
Multiple
- To do XP: Add R2
lm_full <- fit_lm_metamodel(df = df_pa,
x = c("rr", "u_pfs", "u_pd", "c_pfs", "c_pd", "c_thx", "p_pfspd", "p_pfsd", "p_pdd"),
y = "iNMB")
lm_pred_full <- predict(lm_full, data.frame(rr = df_pa$rr,
u_pfs = mean(df_pa$u_pfs),
u_pd = mean(df_pa$u_pd),
c_pfs = mean(df_pa$c_pfs),
c_pd = mean(df_pa$c_pd),
c_thx = mean(df_pa$c_thx),
p_pfspd = mean(df_pa$p_pfspd),
p_pfsd = mean(df_pa$p_pfsd),
p_pdd = mean(df_pa$p_pdd)))
df_obs_pred_full <- data.frame(
Values = df_pa$rr,
Observed = df_pa$iNMB,
Predicted = lm_pred_full
)
ggplot(data = df_obs_pred_full, aes(x = Values, y = Observed)) +
geom_point(shape = 1, colour = "lightgrey") +
geom_line(data = df_obs_pred_full, aes(x = Values, y = Predicted), colour = "blue") +
theme_bw()
plot(lm_full)
df_dsa <- dsa_lm_metamodel(df = df_pa,
lm_metamodel = lm_full)
plot_tornado(df = df_dsa,
df_basecase = df_pa,
outcome = "iNMB")## Loading required package: tidyverse## -- Attaching packages --------------------------------------- tidyverse 1.3.0 --## v tibble 3.0.5 v dplyr 1.0.3
## v tidyr 1.1.2 v stringr 1.4.0
## v readr 1.4.0 v forcats 0.5.1
## v purrr 0.3.4## -- Conflicts ------------------------------------------ tidyverse_conflicts() --
## x readr::col_factor() masks scales::col_factor()
## x purrr::discard() masks scales::discard()
## x dplyr::filter() masks plotly::filter(), stats::filter()
## x dplyr::lag() masks stats::lag()
## x dplyr::select() masks MASS::select(), plotly::select()
5. Add predictions based on metamodel
Other activities/ questions/ to do’s
- Two-way sensitivity analysis based on metamodel.
- Include other functional forms of metamodels
- Include predictions of outcome based on metamodels
- Unit tests of data using
testthat? (automatic check of costs are positive for instance, or utility values between 0-1, sum of transition probabilities not > 1). XAVIER to check whether it has added value to use package or just do it ‘by hand’. Question to Karel: I neem aan dat het mogelijk is om groepen van variabelen te ‘labellen’ (dus alle kosten, utiliteiten, probabiliteiten) en dat de Shiny App het als ‘groep’ ziet?
- Check results of deterministic sensitivity analyses using original model and metamodel.
- XP to do: Add possibility of having multiple scenarios loaded.
- XP to do: Add intermediaire outcomes.
- Shapley value, zou dat kunnen lukken? Zou het mogelijk zijn om dat op de originele data set te doen? KernelExplainer on this github: https://github.com/slundberg/shap. Of is dit iets? https://github.com/nredell/shapFlex, maar ziet eruit dat je ook een model moet fitten (niet model agnostic)
- Check whether violin plot have added value
- Karel, voor selectie variabelen voor de verschillende functies: gebruik van buttons? wat zou handig zijn?